Unveiling Hidden Patterns: Harnessing the Matrix Profile Algorithm for Enhanced Peatland Monitoring and Analysis

Monitoring peatlands involves various scientific methods and technologies, offering crucial insights into the dynamics of these ecosystems, such as carbon storage, hydrological processes, and vegetation composition. By employing advanced monitoring techniques, policymakers, researchers, and land managers can make informed decisions to mitigate threats, restore degraded peatlands, and ensure the sustainable management of these valuable habitats. We present a data analysis and processing approach for distributed monitoring of peatland measurements where hierarchical aggregation and modelling primitives contribute to the robustness of the network by alleviating communication bottlenecks and reducing the energy required for redundant data transmissions. The popular time series data mining method, Matrix Profile (MP), allows efficient dimensionality reduction and feature extraction. The focus is on leveraging the edge computing paradigm to exploit local computing resources and unveil recurring and discordant patterns within monitoring datasets so that only relevant observations are uploaded towards data visualization and decision systems, usually hosted by cloud services. A case study is carried out on real field data for peatland monitoring with outlook on operational and implementation constraints, with discussion accounting for both parametrisation options and underlying seasonal process characteristics.